Conventionally, there have been developed a resource request arbitration device that enables a plurality of arbitration resource request devices such as a processor and a DMA (Direct Memory Access) controller to access a shared resource such as a bus and a unified memory. The resource request devices each issue a resource request to access the shared resource independently from the state of other of the resource request devices. When a plurality of resource request devices simultaneously issue respective resource requests, the resource request arbitration device performs so-called arbitration of setting a priority on each of the resource request devices, and granting access permission to a resource request device having the highest priority among the resource request devices. This prevents collision of accesses to the shared resource.
As basic arbitration methods, there have been known the fixed priority method and the round-robin method. According to the fixed priority method, a priority is fixedly set to each of a plurality of resource request devices. According to the round-robin method, access permission is granted in a cyclic order to a plurality of resource request devices thereby to ensure the fairness of access to a shared resource among the resource request devices.
Also, in order to manage QoS (Quality of Service) represented by a data transfer rate, the maximum queue time, or the like, there has also been known an art of determining a priority for each of a plurality of resource request devices based on progress data of process executed in the resource request device such as a queue time and a slack time of process (see Patent Literature 1 for example).